Face recognition using morphological profile and feature space discriminant analysis

Two face recognition methods based on morphological filters and feature space discriminant analysis (FSDA) are proposed in this paper. Both the proposed methods calculate the morphological profile (MP) of each face sample. The MP contains the contextual information of the face image. Moreover, FSDA, which is a novel feature extraction method introduced in 2015, extracts features with minimum redundant information and maximum class discrimination one. The first proposed method just uses the first component of MP obtained by FSDA, while the second proposed method uses the whole images provided by all opening and closing filters by reconstruction. The dimensionality of each filtered image is reduced by FSDA. Then, the features are fed to a nearest neighbor classifier. Finally the decision fusion rule is used to find the label of each test face image. The experimental results on ORL and Yale face databases show the superior performance of the proposed methods compared to some popular and state-of-the-art face recognition methods.